xref: /netbsd-src/sys/netinet6/ip6_output.c (revision 89c5a767f8fc7a4633b2d409966e2becbb98ff92)
1 /*	$NetBSD: ip6_output.c,v 1.17 2000/03/01 12:49:46 itojun Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*
33  * Copyright (c) 1982, 1986, 1988, 1990, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. All advertising materials mentioning features or use of this software
45  *    must display the following acknowledgement:
46  *	This product includes software developed by the University of
47  *	California, Berkeley and its contributors.
48  * 4. Neither the name of the University nor the names of its contributors
49  *    may be used to endorse or promote products derived from this software
50  *    without specific prior written permission.
51  *
52  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
53  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
56  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62  * SUCH DAMAGE.
63  *
64  *	@(#)ip_output.c	8.3 (Berkeley) 1/21/94
65  */
66 
67 #include "opt_inet.h"
68 #include "opt_ipsec.h"
69 #include "opt_pfil_hooks.h"
70 
71 #include <sys/param.h>
72 #include <sys/malloc.h>
73 #include <sys/mbuf.h>
74 #include <sys/errno.h>
75 #include <sys/protosw.h>
76 #include <sys/socket.h>
77 #include <sys/socketvar.h>
78 #include <sys/systm.h>
79 #include <sys/proc.h>
80 
81 #include <net/if.h>
82 #include <net/route.h>
83 #ifdef PFIL_HOOKS
84 #include <net/pfil.h>
85 #endif
86 
87 #include <netinet/in.h>
88 #include <netinet/in_var.h>
89 #include <netinet/ip6.h>
90 #include <netinet/icmp6.h>
91 #include <netinet6/ip6_var.h>
92 #include <netinet6/in6_pcb.h>
93 #include <netinet6/nd6.h>
94 
95 #ifdef IPSEC
96 #include <netinet6/ipsec.h>
97 #include <netkey/key.h>
98 #include <netkey/key_debug.h>
99 #endif /* IPSEC */
100 
101 #include "loop.h"
102 
103 #include <net/net_osdep.h>
104 
105 #ifdef IPV6FIREWALL
106 #include <netinet6/ip6_fw.h>
107 #endif
108 
109 struct ip6_exthdrs {
110 	struct mbuf *ip6e_ip6;
111 	struct mbuf *ip6e_hbh;
112 	struct mbuf *ip6e_dest1;
113 	struct mbuf *ip6e_rthdr;
114 	struct mbuf *ip6e_dest2;
115 };
116 
117 static int ip6_pcbopts __P((struct ip6_pktopts **, struct mbuf *,
118 			    struct socket *));
119 static int ip6_setmoptions __P((int, struct ip6_moptions **, struct mbuf *));
120 static int ip6_getmoptions __P((int, struct ip6_moptions *, struct mbuf **));
121 static int ip6_copyexthdr __P((struct mbuf **, caddr_t, int));
122 static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
123 				  struct ip6_frag **));
124 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
125 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
126 
127 extern struct ifnet **ifindex2ifnet;
128 extern struct ifnet loif[NLOOP];
129 
130 /*
131  * IP6 output. The packet in mbuf chain m contains a skeletal IP6
132  * header (with pri, len, nxt, hlim, src, dst).
133  * This function may modify ver and hlim only.
134  * The mbuf chain containing the packet will be freed.
135  * The mbuf opt, if present, will not be freed.
136  */
137 int
138 ip6_output(m0, opt, ro, flags, im6o, ifpp)
139 	struct mbuf *m0;
140 	struct ip6_pktopts *opt;
141 	struct route_in6 *ro;
142 	int flags;
143 	struct ip6_moptions *im6o;
144 	struct ifnet **ifpp;		/* XXX: just for statistics */
145 {
146 	struct ip6_hdr *ip6, *mhip6;
147 	struct ifnet *ifp;
148 	struct mbuf *m = m0;
149 	int hlen, tlen, len, off;
150 	struct route_in6 ip6route;
151 	struct sockaddr_in6 *dst;
152 	int error = 0;
153 	struct in6_ifaddr *ia;
154 	u_long mtu;
155 	u_int32_t optlen = 0, plen = 0, unfragpartlen = 0;
156 	struct ip6_exthdrs exthdrs;
157 	struct in6_addr finaldst;
158 	struct route_in6 *ro_pmtu = NULL;
159 	int hdrsplit = 0;
160 	int needipsec = 0;
161 #ifdef PFIL_HOOKS
162 	struct packet_filter_hook *pfh;
163 	struct mbuf *m1;
164 	int rv;
165 #endif /* PFIL_HOOKS */
166 #ifdef IPSEC
167 	int needipsectun = 0;
168 	struct socket *so;
169 	struct secpolicy *sp = NULL;
170 
171 	/* for AH processing. stupid to have "socket" variable in IP layer... */
172 	so = ipsec_getsocket(m);
173 	ipsec_setsocket(m, NULL);
174 	ip6 = mtod(m, struct ip6_hdr *);
175 #endif /* IPSEC */
176 
177 #define MAKE_EXTHDR(hp,mp)						\
178     {									\
179 	if (hp) {							\
180 		struct ip6_ext *eh = (struct ip6_ext *)(hp);		\
181 		error = ip6_copyexthdr((mp), (caddr_t)(hp), 		\
182 				       ((eh)->ip6e_len + 1) << 3);	\
183 		if (error)						\
184 			goto freehdrs;					\
185 	}								\
186     }
187 
188 	bzero(&exthdrs, sizeof(exthdrs));
189 	if (opt) {
190 		/* Hop-by-Hop options header */
191 		MAKE_EXTHDR(opt->ip6po_hbh, &exthdrs.ip6e_hbh);
192 		/* Destination options header(1st part) */
193 		MAKE_EXTHDR(opt->ip6po_dest1, &exthdrs.ip6e_dest1);
194 		/* Routing header */
195 		MAKE_EXTHDR(opt->ip6po_rthdr, &exthdrs.ip6e_rthdr);
196 		/* Destination options header(2nd part) */
197 		MAKE_EXTHDR(opt->ip6po_dest2, &exthdrs.ip6e_dest2);
198 	}
199 
200 #ifdef IPSEC
201 	/* get a security policy for this packet */
202 	if (so == NULL)
203 		sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
204 	else
205 		sp = ipsec6_getpolicybysock(m, IPSEC_DIR_OUTBOUND, so, &error);
206 
207 	if (sp == NULL) {
208 		ipsec6stat.out_inval++;
209 		goto bad;
210 	}
211 
212 	error = 0;
213 
214 	/* check policy */
215 	switch (sp->policy) {
216 	case IPSEC_POLICY_DISCARD:
217 		/*
218 		 * This packet is just discarded.
219 		 */
220 		ipsec6stat.out_polvio++;
221 		goto bad;
222 
223 	case IPSEC_POLICY_BYPASS:
224 	case IPSEC_POLICY_NONE:
225 		/* no need to do IPsec. */
226 		needipsec = 0;
227 		break;
228 
229 	case IPSEC_POLICY_IPSEC:
230 		if (sp->req == NULL) {
231 			/* XXX should be panic ? */
232 			printf("ip6_output: No IPsec request specified.\n");
233 			error = EINVAL;
234 			goto bad;
235 		}
236 		needipsec = 1;
237 		break;
238 
239 	case IPSEC_POLICY_ENTRUST:
240 	default:
241 		printf("ip6_output: Invalid policy found. %d\n", sp->policy);
242 	}
243 #endif /* IPSEC */
244 
245 	/*
246 	 * Calculate the total length of the extension header chain.
247 	 * Keep the length of the unfragmentable part for fragmentation.
248 	 */
249 	optlen = 0;
250 	if (exthdrs.ip6e_hbh) optlen += exthdrs.ip6e_hbh->m_len;
251 	if (exthdrs.ip6e_dest1) optlen += exthdrs.ip6e_dest1->m_len;
252 	if (exthdrs.ip6e_rthdr) optlen += exthdrs.ip6e_rthdr->m_len;
253 	unfragpartlen = optlen + sizeof(struct ip6_hdr);
254 	/* NOTE: we don't add AH/ESP length here. do that later. */
255 	if (exthdrs.ip6e_dest2) optlen += exthdrs.ip6e_dest2->m_len;
256 
257 	/*
258 	 * If we need IPsec, or there is at least one extension header,
259 	 * separate IP6 header from the payload.
260 	 */
261 	if ((needipsec || optlen) && !hdrsplit) {
262 		if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
263 			m = NULL;
264 			goto freehdrs;
265 		}
266 		m = exthdrs.ip6e_ip6;
267 		hdrsplit++;
268 	}
269 
270 	/* adjust pointer */
271 	ip6 = mtod(m, struct ip6_hdr *);
272 
273 	/* adjust mbuf packet header length */
274 	m->m_pkthdr.len += optlen;
275 	plen = m->m_pkthdr.len - sizeof(*ip6);
276 
277 	/* If this is a jumbo payload, insert a jumbo payload option. */
278 	if (plen > IPV6_MAXPACKET) {
279 		if (!hdrsplit) {
280 			if ((error = ip6_splithdr(m, &exthdrs)) != 0) {
281 				m = NULL;
282 				goto freehdrs;
283 			}
284 			m = exthdrs.ip6e_ip6;
285 			hdrsplit++;
286 		}
287 		/* adjust pointer */
288 		ip6 = mtod(m, struct ip6_hdr *);
289 		if ((error = ip6_insert_jumboopt(&exthdrs, plen)) != 0)
290 			goto freehdrs;
291 		ip6->ip6_plen = 0;
292 	} else
293 		ip6->ip6_plen = htons(plen);
294 
295 	/*
296 	 * Concatenate headers and fill in next header fields.
297 	 * Here we have, on "m"
298 	 *	IPv6 payload
299 	 * and we insert headers accordingly.  Finally, we should be getting:
300 	 *	IPv6 hbh dest1 rthdr ah* [esp* dest2 payload]
301 	 *
302 	 * during the header composing process, "m" points to IPv6 header.
303 	 * "mprev" points to an extension header prior to esp.
304 	 */
305 	{
306 		u_char *nexthdrp = &ip6->ip6_nxt;
307 		struct mbuf *mprev = m;
308 
309 		/*
310 		 * we treat dest2 specially.  this makes IPsec processing
311 		 * much easier.
312 		 *
313 		 * result: IPv6 dest2 payload
314 		 * m and mprev will point to IPv6 header.
315 		 */
316 		if (exthdrs.ip6e_dest2) {
317 			if (!hdrsplit)
318 				panic("assumption failed: hdr not split");
319 			exthdrs.ip6e_dest2->m_next = m->m_next;
320 			m->m_next = exthdrs.ip6e_dest2;
321 			*mtod(exthdrs.ip6e_dest2, u_char *) = ip6->ip6_nxt;
322 			ip6->ip6_nxt = IPPROTO_DSTOPTS;
323 		}
324 
325 #define MAKE_CHAIN(m,mp,p,i)\
326     {\
327 	if (m) {\
328 		if (!hdrsplit) \
329 			panic("assumption failed: hdr not split"); \
330 		*mtod((m), u_char *) = *(p);\
331 		*(p) = (i);\
332 		p = mtod((m), u_char *);\
333 		(m)->m_next = (mp)->m_next;\
334 		(mp)->m_next = (m);\
335 		(mp) = (m);\
336 	}\
337     }
338 		/*
339 		 * result: IPv6 hbh dest1 rthdr dest2 payload
340 		 * m will point to IPv6 header.  mprev will point to the
341 		 * extension header prior to dest2 (rthdr in the above case).
342 		 */
343 		MAKE_CHAIN(exthdrs.ip6e_hbh, mprev,
344 			   nexthdrp, IPPROTO_HOPOPTS);
345 		MAKE_CHAIN(exthdrs.ip6e_dest1, mprev,
346 			   nexthdrp, IPPROTO_DSTOPTS);
347 		MAKE_CHAIN(exthdrs.ip6e_rthdr, mprev,
348 			   nexthdrp, IPPROTO_ROUTING);
349 
350 #ifdef IPSEC
351 		if (!needipsec)
352 			goto skip_ipsec2;
353 
354 		/*
355 		 * pointers after IPsec headers are not valid any more.
356 		 * other pointers need a great care too.
357 		 * (IPsec routines should not mangle mbufs prior to AH/ESP)
358 		 */
359 		exthdrs.ip6e_dest2 = NULL;
360 
361 	    {
362 		struct ip6_rthdr *rh = NULL;
363 		int segleft_org = 0;
364 		struct ipsec_output_state state;
365 
366 		if (exthdrs.ip6e_rthdr) {
367 			rh = mtod(exthdrs.ip6e_rthdr, struct ip6_rthdr *);
368 			segleft_org = rh->ip6r_segleft;
369 			rh->ip6r_segleft = 0;
370 		}
371 
372 		bzero(&state, sizeof(state));
373 		state.m = m;
374 		error = ipsec6_output_trans(&state, nexthdrp, mprev, sp, flags,
375 			&needipsectun);
376 		m = state.m;
377 		if (error) {
378 			/* mbuf is already reclaimed in ipsec6_output_trans. */
379 			m = NULL;
380 			switch (error) {
381 			case EHOSTUNREACH:
382 			case ENETUNREACH:
383 			case EMSGSIZE:
384 			case ENOBUFS:
385 			case ENOMEM:
386 				break;
387 			default:
388 				printf("ip6_output (ipsec): error code %d\n", error);
389 				/*fall through*/
390 			case ENOENT:
391 				/* don't show these error codes to the user */
392 				error = 0;
393 				break;
394 			}
395 			goto bad;
396 		}
397 		if (exthdrs.ip6e_rthdr) {
398 			/* ah6_output doesn't modify mbuf chain */
399 			rh->ip6r_segleft = segleft_org;
400 		}
401 	    }
402 skip_ipsec2:;
403 #endif
404 	}
405 
406 	/*
407 	 * If there is a routing header, replace destination address field
408 	 * with the first hop of the routing header.
409 	 */
410 	if (exthdrs.ip6e_rthdr) {
411 		struct ip6_rthdr *rh =
412 			(struct ip6_rthdr *)(mtod(exthdrs.ip6e_rthdr,
413 						  struct ip6_rthdr *));
414 		struct ip6_rthdr0 *rh0;
415 
416 		finaldst = ip6->ip6_dst;
417 		switch(rh->ip6r_type) {
418 		case IPV6_RTHDR_TYPE_0:
419 			 rh0 = (struct ip6_rthdr0 *)rh;
420 			 ip6->ip6_dst = rh0->ip6r0_addr[0];
421 			 bcopy((caddr_t)&rh0->ip6r0_addr[1],
422 				 (caddr_t)&rh0->ip6r0_addr[0],
423 				 sizeof(struct in6_addr)*(rh0->ip6r0_segleft - 1)
424 				 );
425 			 rh0->ip6r0_addr[rh0->ip6r0_segleft - 1] = finaldst;
426 			 break;
427 		default:	/* is it possible? */
428 			 error = EINVAL;
429 			 goto bad;
430 		}
431 	}
432 
433 	/* Source address validation */
434 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src) &&
435 	    (flags & IPV6_DADOUTPUT) == 0) {
436 		error = EOPNOTSUPP;
437 		ip6stat.ip6s_badscope++;
438 		goto bad;
439 	}
440 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src)) {
441 		error = EOPNOTSUPP;
442 		ip6stat.ip6s_badscope++;
443 		goto bad;
444 	}
445 
446 	ip6stat.ip6s_localout++;
447 
448 	/*
449 	 * Route packet.
450 	 */
451 	if (ro == 0) {
452 		ro = &ip6route;
453 		bzero((caddr_t)ro, sizeof(*ro));
454 	}
455 	ro_pmtu = ro;
456 	if (opt && opt->ip6po_rthdr)
457 		ro = &opt->ip6po_route;
458 	dst = (struct sockaddr_in6 *)&ro->ro_dst;
459 	/*
460 	 * If there is a cached route,
461 	 * check that it is to the same destination
462 	 * and is still up. If not, free it and try again.
463 	 */
464 	if (ro->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
465 			 !IN6_ARE_ADDR_EQUAL(&dst->sin6_addr, &ip6->ip6_dst))) {
466 		RTFREE(ro->ro_rt);
467 		ro->ro_rt = (struct rtentry *)0;
468 	}
469 	if (ro->ro_rt == 0) {
470 		bzero(dst, sizeof(*dst));
471 		dst->sin6_family = AF_INET6;
472 		dst->sin6_len = sizeof(struct sockaddr_in6);
473 		dst->sin6_addr = ip6->ip6_dst;
474 	}
475 #ifdef IPSEC
476 	if (needipsec && needipsectun) {
477 		struct ipsec_output_state state;
478 
479 		/*
480 		 * All the extension headers will become inaccessible
481 		 * (since they can be encrypted).
482 		 * Don't panic, we need no more updates to extension headers
483 		 * on inner IPv6 packet (since they are now encapsulated).
484 		 *
485 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
486 		 */
487 		bzero(&exthdrs, sizeof(exthdrs));
488 		exthdrs.ip6e_ip6 = m;
489 
490 		bzero(&state, sizeof(state));
491 		state.m = m;
492 		state.ro = (struct route *)ro;
493 		state.dst = (struct sockaddr *)dst;
494 
495 		error = ipsec6_output_tunnel(&state, sp, flags);
496 
497 		m = state.m;
498 		ro = (struct route_in6 *)state.ro;
499 		dst = (struct sockaddr_in6 *)state.dst;
500 		if (error) {
501 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
502 			m0 = m = NULL;
503 			m = NULL;
504 			switch (error) {
505 			case EHOSTUNREACH:
506 			case ENETUNREACH:
507 			case EMSGSIZE:
508 			case ENOBUFS:
509 			case ENOMEM:
510 				break;
511 			default:
512 				printf("ip6_output (ipsec): error code %d\n", error);
513 				/*fall through*/
514 			case ENOENT:
515 				/* don't show these error codes to the user */
516 				error = 0;
517 				break;
518 			}
519 			goto bad;
520 		}
521 
522 		exthdrs.ip6e_ip6 = m;
523 	}
524 #endif /*IPESC*/
525 
526 	if (!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
527 		/* Unicast */
528 
529 #define ifatoia6(ifa)	((struct in6_ifaddr *)(ifa))
530 #define sin6tosa(sin6)	((struct sockaddr *)(sin6))
531 		/* xxx
532 		 * interface selection comes here
533 		 * if an interface is specified from an upper layer,
534 		 * ifp must point it.
535 		 */
536 		if (ro->ro_rt == 0) {
537 			/*
538 			 * NetBSD/OpenBSD always clones routes, if parent is
539 			 * PRF_CLONING.
540 			 */
541 			rtalloc((struct route *)ro);
542 		}
543 		if (ro->ro_rt == 0) {
544 			ip6stat.ip6s_noroute++;
545 			error = EHOSTUNREACH;
546 			/* XXX in6_ifstat_inc(ifp, ifs6_out_discard); */
547 			goto bad;
548 		}
549 		ia = ifatoia6(ro->ro_rt->rt_ifa);
550 		ifp = ro->ro_rt->rt_ifp;
551 		ro->ro_rt->rt_use++;
552 		if (ro->ro_rt->rt_flags & RTF_GATEWAY)
553 			dst = (struct sockaddr_in6 *)ro->ro_rt->rt_gateway;
554 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
555 
556 		in6_ifstat_inc(ifp, ifs6_out_request);
557 
558 		/*
559 		 * Check if there is the outgoing interface conflicts with
560 		 * the interface specified by ifi6_ifindex(if specified).
561 		 * Note that loopback interface is always okay.
562 		 * (this happens when we are sending packet toward my
563 		 * interface)
564 		 */
565 		if (opt && opt->ip6po_pktinfo
566 		 && opt->ip6po_pktinfo->ipi6_ifindex) {
567 			if (!(ifp->if_flags & IFF_LOOPBACK)
568 			 && ifp->if_index != opt->ip6po_pktinfo->ipi6_ifindex) {
569 				ip6stat.ip6s_noroute++;
570 				in6_ifstat_inc(ifp, ifs6_out_discard);
571 				error = EHOSTUNREACH;
572 				goto bad;
573 			}
574 		}
575 
576 		if (opt && opt->ip6po_hlim != -1)
577 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
578 	} else {
579 		/* Multicast */
580 		struct	in6_multi *in6m;
581 
582 		m->m_flags = (m->m_flags & ~M_BCAST) | M_MCAST;
583 
584 		/*
585 		 * See if the caller provided any multicast options
586 		 */
587 		ifp = NULL;
588 		if (im6o != NULL) {
589 			ip6->ip6_hlim = im6o->im6o_multicast_hlim;
590 			if (im6o->im6o_multicast_ifp != NULL)
591 				ifp = im6o->im6o_multicast_ifp;
592 		} else
593 			ip6->ip6_hlim = ip6_defmcasthlim;
594 
595 		/*
596 		 * See if the caller provided the outgoing interface
597 		 * as an ancillary data.
598 		 * Boundary check for ifindex is assumed to be already done.
599 		 */
600 		if (opt && opt->ip6po_pktinfo && opt->ip6po_pktinfo->ipi6_ifindex)
601 			ifp = ifindex2ifnet[opt->ip6po_pktinfo->ipi6_ifindex];
602 
603 		/*
604 		 * If the destination is a node-local scope multicast,
605 		 * the packet should be loop-backed only.
606 		 */
607 		if (IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst)) {
608 			/*
609 			 * If the outgoing interface is already specified,
610 			 * it should be a loopback interface.
611 			 */
612 			if (ifp && (ifp->if_flags & IFF_LOOPBACK) == 0) {
613 				ip6stat.ip6s_badscope++;
614 				error = ENETUNREACH; /* XXX: better error? */
615 				/* XXX correct ifp? */
616 				in6_ifstat_inc(ifp, ifs6_out_discard);
617 				goto bad;
618 			}
619 			else {
620 				ifp = &loif[0];
621 			}
622 		}
623 
624 		if (opt && opt->ip6po_hlim != -1)
625 			ip6->ip6_hlim = opt->ip6po_hlim & 0xff;
626 
627 		/*
628 		 * If caller did not provide an interface lookup a
629 		 * default in the routing table.  This is either a
630 		 * default for the speicfied group (i.e. a host
631 		 * route), or a multicast default (a route for the
632 		 * ``net'' ff00::/8).
633 		 */
634 		if (ifp == NULL) {
635 			if (ro->ro_rt == 0) {
636 				ro->ro_rt = rtalloc1((struct sockaddr *)
637 						&ro->ro_dst, 0
638 						);
639 			}
640 			if (ro->ro_rt == 0) {
641 				ip6stat.ip6s_noroute++;
642 				error = EHOSTUNREACH;
643 				/* XXX in6_ifstat_inc(ifp, ifs6_out_discard) */
644 				goto bad;
645 			}
646 			ia = ifatoia6(ro->ro_rt->rt_ifa);
647 			ifp = ro->ro_rt->rt_ifp;
648 			ro->ro_rt->rt_use++;
649 		}
650 
651 		if ((flags & IPV6_FORWARDING) == 0)
652 			in6_ifstat_inc(ifp, ifs6_out_request);
653 		in6_ifstat_inc(ifp, ifs6_out_mcast);
654 
655 		/*
656 		 * Confirm that the outgoing interface supports multicast.
657 		 */
658 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
659 			ip6stat.ip6s_noroute++;
660 			in6_ifstat_inc(ifp, ifs6_out_discard);
661 			error = ENETUNREACH;
662 			goto bad;
663 		}
664 		IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
665 		if (in6m != NULL &&
666 		   (im6o == NULL || im6o->im6o_multicast_loop)) {
667 			/*
668 			 * If we belong to the destination multicast group
669 			 * on the outgoing interface, and the caller did not
670 			 * forbid loopback, loop back a copy.
671 			 */
672 			ip6_mloopback(ifp, m, dst);
673 		} else {
674 			/*
675 			 * If we are acting as a multicast router, perform
676 			 * multicast forwarding as if the packet had just
677 			 * arrived on the interface to which we are about
678 			 * to send.  The multicast forwarding function
679 			 * recursively calls this function, using the
680 			 * IPV6_FORWARDING flag to prevent infinite recursion.
681 			 *
682 			 * Multicasts that are looped back by ip6_mloopback(),
683 			 * above, will be forwarded by the ip6_input() routine,
684 			 * if necessary.
685 			 */
686 			if (ip6_mrouter && (flags & IPV6_FORWARDING) == 0) {
687 				if (ip6_mforward(ip6, ifp, m) != NULL) {
688 					m_freem(m);
689 					goto done;
690 				}
691 			}
692 		}
693 		/*
694 		 * Multicasts with a hoplimit of zero may be looped back,
695 		 * above, but must not be transmitted on a network.
696 		 * Also, multicasts addressed to the loopback interface
697 		 * are not sent -- the above call to ip6_mloopback() will
698 		 * loop back a copy if this host actually belongs to the
699 		 * destination group on the loopback interface.
700 		 */
701 		if (ip6->ip6_hlim == 0 || (ifp->if_flags & IFF_LOOPBACK)) {
702 			m_freem(m);
703 			goto done;
704 		}
705 	}
706 
707 	/*
708 	 * Fill the outgoing inteface to tell the upper layer
709 	 * to increment per-interface statistics.
710 	 */
711 	if (ifpp)
712 		*ifpp = ifp;
713 
714 	/*
715 	 * Determine path MTU.
716 	 */
717 	if (ro_pmtu != ro) {
718 		/* The first hop and the final destination may differ. */
719 		struct sockaddr_in6 *sin6_fin =
720 			(struct sockaddr_in6 *)&ro_pmtu->ro_dst;
721 		if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
722 				       !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
723 							   &finaldst))) {
724 			RTFREE(ro_pmtu->ro_rt);
725 			ro_pmtu->ro_rt = (struct rtentry *)0;
726 		}
727 		if (ro_pmtu->ro_rt == 0) {
728 			bzero(sin6_fin, sizeof(*sin6_fin));
729 			sin6_fin->sin6_family = AF_INET6;
730 			sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
731 			sin6_fin->sin6_addr = finaldst;
732 
733 			rtalloc((struct route *)ro_pmtu);
734 		}
735 	}
736 	if (ro_pmtu->ro_rt != NULL) {
737 		u_int32_t ifmtu = nd_ifinfo[ifp->if_index].linkmtu;
738 
739 		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
740 		if (mtu > ifmtu) {
741 			/*
742 			 * The MTU on the route is larger than the MTU on
743 			 * the interface!  This shouldn't happen, unless the
744 			 * MTU of the interface has been changed after the
745 			 * interface was brought up.  Change the MTU in the
746 			 * route to match the interface MTU (as long as the
747 			 * field isn't locked).
748 			 */
749 			 mtu = ifmtu;
750 			 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
751 				 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
752 		}
753 	} else {
754 		mtu = nd_ifinfo[ifp->if_index].linkmtu;
755 	}
756 
757 	/*
758 	 * Fake link-local scope-class addresses
759 	 */
760 	if ((ifp->if_flags & IFF_LOOPBACK) == 0) {
761 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
762 			ip6->ip6_src.s6_addr16[1] = 0;
763 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
764 			ip6->ip6_dst.s6_addr16[1] = 0;
765 	}
766 
767 	/*
768 	 * If the outgoing packet contains a hop-by-hop options header,
769 	 * it must be examined and processed even by the source node.
770 	 * (RFC 2460, section 4.)
771 	 */
772 	if (exthdrs.ip6e_hbh) {
773 		struct ip6_hbh *hbh = mtod(exthdrs.ip6e_hbh,
774 					   struct ip6_hbh *);
775 		u_int32_t dummy1; /* XXX unused */
776 		u_int32_t dummy2; /* XXX unused */
777 
778 		/*
779 		 *  XXX: if we have to send an ICMPv6 error to the sender,
780 		 *       we need the M_LOOP flag since icmp6_error() expects
781 		 *       the IPv6 and the hop-by-hop options header are
782 		 *       continuous unless the flag is set.
783 		 */
784 		m->m_flags |= M_LOOP;
785 		m->m_pkthdr.rcvif = ifp;
786 		if (ip6_process_hopopts(m,
787 					(u_int8_t *)(hbh + 1),
788 					((hbh->ip6h_len + 1) << 3) -
789 					sizeof(struct ip6_hbh),
790 					&dummy1, &dummy2) < 0) {
791 			/* m was already freed at this point */
792 			error = EINVAL;/* better error? */
793 			goto done;
794 		}
795 		m->m_flags &= ~M_LOOP; /* XXX */
796 		m->m_pkthdr.rcvif = NULL;
797 	}
798 
799 #ifdef PFIL_HOOKS
800 	/*
801 	 * Run through list of hooks for output packets.
802 	 */
803 	m1 = m;
804 	pfh = pfil_hook_get(PFIL_OUT, &inetsw[ip_protox[IPPROTO_IPV6]].pr_pfh);
805 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
806 		if (pfh->pfil_func) {
807 		    	rv = pfh->pfil_func(ip6, sizeof(*ip6), ifp, 1, &m1);
808 			if (rv) {
809 				error = EHOSTUNREACH;
810 				goto done;
811 			}
812 			m = m1;
813 			if (m == NULL)
814 				goto done;
815 			ip6 = mtod(m, struct ip6_hdr *);
816 		}
817 #endif /* PFIL_HOOKS */
818 	/*
819 	 * Send the packet to the outgoing interface.
820 	 * If necessary, do IPv6 fragmentation before sending.
821 	 */
822 	tlen = m->m_pkthdr.len;
823 	if (tlen <= mtu
824 #ifdef notyet
825 	    /*
826 	     * On any link that cannot convey a 1280-octet packet in one piece,
827 	     * link-specific fragmentation and reassembly must be provided at
828 	     * a layer below IPv6. [RFC 2460, sec.5]
829 	     * Thus if the interface has ability of link-level fragmentation,
830 	     * we can just send the packet even if the packet size is
831 	     * larger than the link's MTU.
832 	     * XXX: IFF_FRAGMENTABLE (or such) flag has not been defined yet...
833 	     */
834 
835 	    || ifp->if_flags & IFF_FRAGMENTABLE
836 #endif
837 	    )
838 	{
839 #ifdef IFA_STATS
840 		if (IFA_STATS) {
841 			struct in6_ifaddr *ia6;
842 			ip6 = mtod(m, struct ip6_hdr *);
843 			ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
844 			if (ia6) {
845 				ia->ia_ifa.ifa_data.ifad_outbytes +=
846 					m->m_pkthdr.len;
847 			}
848 		}
849 #endif
850 #ifdef OLDIP6OUTPUT
851 		error = (*ifp->if_output)(ifp, m, (struct sockaddr *)dst,
852 					  ro->ro_rt);
853 #else
854 		error = nd6_output(ifp, m, dst, ro->ro_rt);
855 #endif
856 		goto done;
857 	} else if (mtu < IPV6_MMTU) {
858 		/*
859 		 * note that path MTU is never less than IPV6_MMTU
860 		 * (see icmp6_input).
861 		 */
862 		error = EMSGSIZE;
863 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
864 		goto bad;
865 	} else if (ip6->ip6_plen == 0) { /* jumbo payload cannot be fragmented */
866 		error = EMSGSIZE;
867 		in6_ifstat_inc(ifp, ifs6_out_fragfail);
868 		goto bad;
869 	} else {
870 		struct mbuf **mnext, *m_frgpart;
871 		struct ip6_frag *ip6f;
872 		u_int32_t id = htonl(ip6_id++);
873 		u_char nextproto;
874 
875 		/*
876 		 * Too large for the destination or interface;
877 		 * fragment if possible.
878 		 * Must be able to put at least 8 bytes per fragment.
879 		 */
880 		hlen = unfragpartlen;
881 		if (mtu > IPV6_MAXPACKET)
882 			mtu = IPV6_MAXPACKET;
883 		len = (mtu - hlen - sizeof(struct ip6_frag)) & ~7;
884 		if (len < 8) {
885 			error = EMSGSIZE;
886 			in6_ifstat_inc(ifp, ifs6_out_fragfail);
887 			goto bad;
888 		}
889 
890 		mnext = &m->m_nextpkt;
891 
892 		/*
893 		 * Change the next header field of the last header in the
894 		 * unfragmentable part.
895 		 */
896 		if (exthdrs.ip6e_rthdr) {
897 			nextproto = *mtod(exthdrs.ip6e_rthdr, u_char *);
898 			*mtod(exthdrs.ip6e_rthdr, u_char *) = IPPROTO_FRAGMENT;
899 		}
900 		else if (exthdrs.ip6e_dest1) {
901 			nextproto = *mtod(exthdrs.ip6e_dest1, u_char *);
902 			*mtod(exthdrs.ip6e_dest1, u_char *) = IPPROTO_FRAGMENT;
903 		}
904 		else if (exthdrs.ip6e_hbh) {
905 			nextproto = *mtod(exthdrs.ip6e_hbh, u_char *);
906 			*mtod(exthdrs.ip6e_hbh, u_char *) = IPPROTO_FRAGMENT;
907 		}
908 		else {
909 			nextproto = ip6->ip6_nxt;
910 			ip6->ip6_nxt = IPPROTO_FRAGMENT;
911 		}
912 
913 		/*
914 		 * Loop through length of segment after first fragment,
915 		 * make new header and copy data of each part and link onto chain.
916 		 */
917 		m0 = m;
918 		for (off = hlen; off < tlen; off += len) {
919 			MGETHDR(m, M_DONTWAIT, MT_HEADER);
920 			if (!m) {
921 				error = ENOBUFS;
922 				ip6stat.ip6s_odropped++;
923 				goto sendorfree;
924 			}
925 			m->m_flags = m0->m_flags & M_COPYFLAGS;
926 			*mnext = m;
927 			mnext = &m->m_nextpkt;
928 			m->m_data += max_linkhdr;
929 			mhip6 = mtod(m, struct ip6_hdr *);
930 			*mhip6 = *ip6;
931 			m->m_len = sizeof(*mhip6);
932  			error = ip6_insertfraghdr(m0, m, hlen, &ip6f);
933  			if (error) {
934 				ip6stat.ip6s_odropped++;
935 				goto sendorfree;
936 			}
937 			ip6f->ip6f_offlg = htons((u_short)((off - hlen) & ~7));
938 			if (off + len >= tlen)
939 				len = tlen - off;
940 			else
941 				ip6f->ip6f_offlg |= IP6F_MORE_FRAG;
942 			mhip6->ip6_plen = htons((u_short)(len + hlen +
943 							  sizeof(*ip6f) -
944 							  sizeof(struct ip6_hdr)));
945 			if ((m_frgpart = m_copy(m0, off, len)) == 0) {
946 				error = ENOBUFS;
947 				ip6stat.ip6s_odropped++;
948 				goto sendorfree;
949 			}
950 			m_cat(m, m_frgpart);
951 			m->m_pkthdr.len = len + hlen + sizeof(*ip6f);
952 			m->m_pkthdr.rcvif = (struct ifnet *)0;
953 			ip6f->ip6f_reserved = 0;
954 			ip6f->ip6f_ident = id;
955 			ip6f->ip6f_nxt = nextproto;
956 			ip6stat.ip6s_ofragments++;
957 			in6_ifstat_inc(ifp, ifs6_out_fragcreat);
958 		}
959 
960 		in6_ifstat_inc(ifp, ifs6_out_fragok);
961 	}
962 
963 	/*
964 	 * Remove leading garbages.
965 	 */
966 sendorfree:
967 	m = m0->m_nextpkt;
968 	m0->m_nextpkt = 0;
969 	m_freem(m0);
970 	for (m0 = m; m; m = m0) {
971 		m0 = m->m_nextpkt;
972 		m->m_nextpkt = 0;
973 		if (error == 0) {
974 #ifdef IFA_STATS
975 			if (IFA_STATS) {
976 				struct in6_ifaddr *ia6;
977 				ip6 = mtod(m, struct ip6_hdr *);
978 				ia6 = in6_ifawithifp(ifp, &ip6->ip6_src);
979 				if (ia6) {
980 					ia->ia_ifa.ifa_data.ifad_outbytes +=
981 						m->m_pkthdr.len;
982 				}
983 			}
984 #endif
985 #ifdef OLDIP6OUTPUT
986 			error = (*ifp->if_output)(ifp, m,
987 						  (struct sockaddr *)dst,
988 						  ro->ro_rt);
989 #else
990 			error = nd6_output(ifp, m, dst, ro->ro_rt);
991 #endif
992 		}
993 		else
994 			m_freem(m);
995 	}
996 
997 	if (error == 0)
998 		ip6stat.ip6s_fragmented++;
999 
1000 done:
1001 	if (ro == &ip6route && ro->ro_rt) { /* brace necessary for RTFREE */
1002 		RTFREE(ro->ro_rt);
1003 	} else if (ro_pmtu == &ip6route && ro_pmtu->ro_rt) {
1004 		RTFREE(ro_pmtu->ro_rt);
1005 	}
1006 
1007 #ifdef IPSEC
1008 	if (sp != NULL)
1009 		key_freesp(sp);
1010 #endif /* IPSEC */
1011 
1012 	return(error);
1013 
1014 freehdrs:
1015 	m_freem(exthdrs.ip6e_hbh);	/* m_freem will check if mbuf is 0 */
1016 	m_freem(exthdrs.ip6e_dest1);
1017 	m_freem(exthdrs.ip6e_rthdr);
1018 	m_freem(exthdrs.ip6e_dest2);
1019 	/* fall through */
1020 bad:
1021 	m_freem(m);
1022 	goto done;
1023 }
1024 
1025 static int
1026 ip6_copyexthdr(mp, hdr, hlen)
1027 	struct mbuf **mp;
1028 	caddr_t hdr;
1029 	int hlen;
1030 {
1031 	struct mbuf *m;
1032 
1033 	if (hlen > MCLBYTES)
1034 		return(ENOBUFS); /* XXX */
1035 
1036 	MGET(m, M_DONTWAIT, MT_DATA);
1037 	if (!m)
1038 		return(ENOBUFS);
1039 
1040 	if (hlen > MLEN) {
1041 		MCLGET(m, M_DONTWAIT);
1042 		if ((m->m_flags & M_EXT) == 0) {
1043 			m_free(m);
1044 			return(ENOBUFS);
1045 		}
1046 	}
1047 	m->m_len = hlen;
1048 	if (hdr)
1049 		bcopy(hdr, mtod(m, caddr_t), hlen);
1050 
1051 	*mp = m;
1052 	return(0);
1053 }
1054 
1055 /*
1056  * Insert jumbo payload option.
1057  */
1058 static int
1059 ip6_insert_jumboopt(exthdrs, plen)
1060 	struct ip6_exthdrs *exthdrs;
1061 	u_int32_t plen;
1062 {
1063 	struct mbuf *mopt;
1064 	u_char *optbuf;
1065 
1066 #define JUMBOOPTLEN	8	/* length of jumbo payload option and padding */
1067 
1068 	/*
1069 	 * If there is no hop-by-hop options header, allocate new one.
1070 	 * If there is one but it doesn't have enough space to store the
1071 	 * jumbo payload option, allocate a cluster to store the whole options.
1072 	 * Otherwise, use it to store the options.
1073 	 */
1074 	if (exthdrs->ip6e_hbh == 0) {
1075 		MGET(mopt, M_DONTWAIT, MT_DATA);
1076 		if (mopt == 0)
1077 			return(ENOBUFS);
1078 		mopt->m_len = JUMBOOPTLEN;
1079 		optbuf = mtod(mopt, u_char *);
1080 		optbuf[1] = 0;	/* = ((JUMBOOPTLEN) >> 3) - 1 */
1081 		exthdrs->ip6e_hbh = mopt;
1082 	}
1083 	else {
1084 		struct ip6_hbh *hbh;
1085 
1086 		mopt = exthdrs->ip6e_hbh;
1087 		if (M_TRAILINGSPACE(mopt) < JUMBOOPTLEN) {
1088 			caddr_t oldoptp = mtod(mopt, caddr_t);
1089 			int oldoptlen = mopt->m_len;
1090 
1091 			if (mopt->m_flags & M_EXT)
1092 				return(ENOBUFS); /* XXX */
1093 			MCLGET(mopt, M_DONTWAIT);
1094 			if ((mopt->m_flags & M_EXT) == 0)
1095 				return(ENOBUFS);
1096 
1097 			bcopy(oldoptp, mtod(mopt, caddr_t), oldoptlen);
1098 			optbuf = mtod(mopt, caddr_t) + oldoptlen;
1099 			mopt->m_len = oldoptlen + JUMBOOPTLEN;
1100 		}
1101 		else {
1102 			optbuf = mtod(mopt, u_char *) + mopt->m_len;
1103 			mopt->m_len += JUMBOOPTLEN;
1104 		}
1105 		optbuf[0] = IP6OPT_PADN;
1106 		optbuf[1] = 1;
1107 
1108 		/*
1109 		 * Adjust the header length according to the pad and
1110 		 * the jumbo payload option.
1111 		 */
1112 		hbh = mtod(mopt, struct ip6_hbh *);
1113 		hbh->ip6h_len += (JUMBOOPTLEN >> 3);
1114 	}
1115 
1116 	/* fill in the option. */
1117 	optbuf[2] = IP6OPT_JUMBO;
1118 	optbuf[3] = 4;
1119 	*(u_int32_t *)&optbuf[4] = htonl(plen + JUMBOOPTLEN);
1120 
1121 	/* finally, adjust the packet header length */
1122 	exthdrs->ip6e_ip6->m_pkthdr.len += JUMBOOPTLEN;
1123 
1124 	return(0);
1125 #undef JUMBOOPTLEN
1126 }
1127 
1128 /*
1129  * Insert fragment header and copy unfragmentable header portions.
1130  */
1131 static int
1132 ip6_insertfraghdr(m0, m, hlen, frghdrp)
1133 	struct mbuf *m0, *m;
1134 	int hlen;
1135 	struct ip6_frag **frghdrp;
1136 {
1137 	struct mbuf *n, *mlast;
1138 
1139 	if (hlen > sizeof(struct ip6_hdr)) {
1140 		n = m_copym(m0, sizeof(struct ip6_hdr),
1141 			    hlen - sizeof(struct ip6_hdr), M_DONTWAIT);
1142 		if (n == 0)
1143 			return(ENOBUFS);
1144 		m->m_next = n;
1145 	}
1146 	else
1147 		n = m;
1148 
1149 	/* Search for the last mbuf of unfragmentable part. */
1150 	for (mlast = n; mlast->m_next; mlast = mlast->m_next)
1151 		;
1152 
1153 	if ((mlast->m_flags & M_EXT) == 0 &&
1154 	    M_TRAILINGSPACE(mlast) < sizeof(struct ip6_frag)) {
1155 		/* use the trailing space of the last mbuf for the fragment hdr */
1156 		*frghdrp =
1157 			(struct ip6_frag *)(mtod(mlast, caddr_t) + mlast->m_len);
1158 		mlast->m_len += sizeof(struct ip6_frag);
1159 		m->m_pkthdr.len += sizeof(struct ip6_frag);
1160 	}
1161 	else {
1162 		/* allocate a new mbuf for the fragment header */
1163 		struct mbuf *mfrg;
1164 
1165 		MGET(mfrg, M_DONTWAIT, MT_DATA);
1166 		if (mfrg == 0)
1167 			return(ENOBUFS);
1168 		mfrg->m_len = sizeof(struct ip6_frag);
1169 		*frghdrp = mtod(mfrg, struct ip6_frag *);
1170 		mlast->m_next = mfrg;
1171 	}
1172 
1173 	return(0);
1174 }
1175 
1176 /*
1177  * IP6 socket option processing.
1178  */
1179 int
1180 ip6_ctloutput(op, so, level, optname, mp)
1181 	int op;
1182 	struct socket *so;
1183 	int level, optname;
1184 	struct mbuf **mp;
1185 {
1186 	register struct in6pcb *in6p = sotoin6pcb(so);
1187 	register struct mbuf *m = *mp;
1188 	register int optval = 0;
1189 	int error = 0;
1190 	struct proc *p = curproc;	/* XXX */
1191 
1192 	if (level == IPPROTO_IPV6)
1193 		switch (op) {
1194 
1195 		case PRCO_SETOPT:
1196 			switch (optname) {
1197 			case IPV6_PKTOPTIONS:
1198 				return(ip6_pcbopts(&in6p->in6p_outputopts,
1199 						   m, so));
1200 			case IPV6_HOPOPTS:
1201 			case IPV6_DSTOPTS:
1202 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
1203 					error = EPERM;
1204 					break;
1205 				}
1206 				/* fall through */
1207 			case IPV6_UNICAST_HOPS:
1208 			case IPV6_RECVOPTS:
1209 			case IPV6_RECVRETOPTS:
1210 			case IPV6_RECVDSTADDR:
1211 			case IPV6_PKTINFO:
1212 			case IPV6_HOPLIMIT:
1213 			case IPV6_RTHDR:
1214 			case IPV6_CHECKSUM:
1215 			case IPV6_FAITH:
1216 #ifndef INET6_BINDV6ONLY
1217 			case IPV6_BINDV6ONLY:
1218 #endif
1219 				if (!m || m->m_len != sizeof(int))
1220 					error = EINVAL;
1221 				else {
1222 					optval = *mtod(m, int *);
1223 					switch (optname) {
1224 
1225 					case IPV6_UNICAST_HOPS:
1226 						if (optval < -1 || optval >= 256)
1227 							error = EINVAL;
1228 						else {
1229 							/* -1 = kernel default */
1230 							in6p->in6p_hops = optval;
1231 						}
1232 						break;
1233 #define OPTSET(bit) \
1234 	if (optval) \
1235 		in6p->in6p_flags |= bit; \
1236 	else \
1237 		in6p->in6p_flags &= ~bit;
1238 
1239 					case IPV6_RECVOPTS:
1240 						OPTSET(IN6P_RECVOPTS);
1241 						break;
1242 
1243 					case IPV6_RECVRETOPTS:
1244 						OPTSET(IN6P_RECVRETOPTS);
1245 						break;
1246 
1247 					case IPV6_RECVDSTADDR:
1248 						OPTSET(IN6P_RECVDSTADDR);
1249 						break;
1250 
1251 					case IPV6_PKTINFO:
1252 						OPTSET(IN6P_PKTINFO);
1253 						break;
1254 
1255 					case IPV6_HOPLIMIT:
1256 						OPTSET(IN6P_HOPLIMIT);
1257 						break;
1258 
1259 					case IPV6_HOPOPTS:
1260 						OPTSET(IN6P_HOPOPTS);
1261 						break;
1262 
1263 					case IPV6_DSTOPTS:
1264 						OPTSET(IN6P_DSTOPTS);
1265 						break;
1266 
1267 					case IPV6_RTHDR:
1268 						OPTSET(IN6P_RTHDR);
1269 						break;
1270 
1271 					case IPV6_CHECKSUM:
1272 						in6p->in6p_cksum = optval;
1273 						break;
1274 
1275 					case IPV6_FAITH:
1276 						OPTSET(IN6P_FAITH);
1277 						break;
1278 
1279 #ifndef INET6_BINDV6ONLY
1280 					case IPV6_BINDV6ONLY:
1281 						OPTSET(IN6P_BINDV6ONLY);
1282 						break;
1283 #endif
1284 					}
1285 				}
1286 				break;
1287 #undef OPTSET
1288 
1289 			case IPV6_MULTICAST_IF:
1290 			case IPV6_MULTICAST_HOPS:
1291 			case IPV6_MULTICAST_LOOP:
1292 			case IPV6_JOIN_GROUP:
1293 			case IPV6_LEAVE_GROUP:
1294 				error =	ip6_setmoptions(optname, &in6p->in6p_moptions, m);
1295 				break;
1296 
1297 			case IPV6_PORTRANGE:
1298 				optval = *mtod(m, int *);
1299 
1300 				switch (optval) {
1301 				case IPV6_PORTRANGE_DEFAULT:
1302 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1303 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1304 					break;
1305 
1306 				case IPV6_PORTRANGE_HIGH:
1307 					in6p->in6p_flags &= ~(IN6P_LOWPORT);
1308 					in6p->in6p_flags |= IN6P_HIGHPORT;
1309 					break;
1310 
1311 				case IPV6_PORTRANGE_LOW:
1312 					in6p->in6p_flags &= ~(IN6P_HIGHPORT);
1313 					in6p->in6p_flags |= IN6P_LOWPORT;
1314 					break;
1315 
1316 				default:
1317 					error = EINVAL;
1318 					break;
1319 				}
1320 				break;
1321 
1322 #ifdef IPSEC
1323 			case IPV6_IPSEC_POLICY:
1324 			    {
1325 				caddr_t req = NULL;
1326 				size_t len = 0;
1327 
1328 				int priv = 0;
1329 				if (p == 0 || suser(p->p_ucred, &p->p_acflag))
1330 					priv = 0;
1331 				else
1332 					priv = 1;
1333 				if (m) {
1334 					req = mtod(m, caddr_t);
1335 					len = m->m_len;
1336 				}
1337 				error = ipsec6_set_policy(in6p,
1338 				                   optname, req, len, priv);
1339 			    }
1340 				break;
1341 #endif /* IPSEC */
1342 
1343 			default:
1344 				error = ENOPROTOOPT;
1345 				break;
1346 			}
1347 			if (m)
1348 				(void)m_free(m);
1349 			break;
1350 
1351 		case PRCO_GETOPT:
1352 			switch (optname) {
1353 
1354 			case IPV6_OPTIONS:
1355 			case IPV6_RETOPTS:
1356 #if 0
1357 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1358 				if (in6p->in6p_options) {
1359 					m->m_len = in6p->in6p_options->m_len;
1360 					bcopy(mtod(in6p->in6p_options, caddr_t),
1361 					      mtod(m, caddr_t),
1362 					      (unsigned)m->m_len);
1363 				} else
1364 					m->m_len = 0;
1365 				break;
1366 #else
1367 				error = ENOPROTOOPT;
1368 				break;
1369 #endif
1370 
1371 			case IPV6_PKTOPTIONS:
1372 				if (in6p->in6p_options) {
1373 					*mp = m_copym(in6p->in6p_options, 0,
1374 						      M_COPYALL, M_WAIT);
1375 				} else {
1376 					*mp = m_get(M_WAIT, MT_SOOPTS);
1377 					(*mp)->m_len = 0;
1378 				}
1379 				break;
1380 
1381 			case IPV6_HOPOPTS:
1382 			case IPV6_DSTOPTS:
1383 				if (p == 0 || suser(p->p_ucred, &p->p_acflag)) {
1384 					error = EPERM;
1385 					break;
1386 				}
1387 				/* fall through */
1388 			case IPV6_UNICAST_HOPS:
1389 			case IPV6_RECVOPTS:
1390 			case IPV6_RECVRETOPTS:
1391 			case IPV6_RECVDSTADDR:
1392 			case IPV6_PORTRANGE:
1393 			case IPV6_PKTINFO:
1394 			case IPV6_HOPLIMIT:
1395 			case IPV6_RTHDR:
1396 			case IPV6_CHECKSUM:
1397 			case IPV6_FAITH:
1398 #ifndef INET6_BINDV6ONLY
1399 			case IPV6_BINDV6ONLY:
1400 #endif
1401 				*mp = m = m_get(M_WAIT, MT_SOOPTS);
1402 				m->m_len = sizeof(int);
1403 				switch (optname) {
1404 
1405 				case IPV6_UNICAST_HOPS:
1406 					optval = in6p->in6p_hops;
1407 					break;
1408 
1409 #define OPTBIT(bit) (in6p->in6p_flags & bit ? 1 : 0)
1410 
1411 				case IPV6_RECVOPTS:
1412 					optval = OPTBIT(IN6P_RECVOPTS);
1413 					break;
1414 
1415 				case IPV6_RECVRETOPTS:
1416 					optval = OPTBIT(IN6P_RECVRETOPTS);
1417 					break;
1418 
1419 				case IPV6_RECVDSTADDR:
1420 					optval = OPTBIT(IN6P_RECVDSTADDR);
1421 					break;
1422 
1423 				case IPV6_PORTRANGE:
1424 				    {
1425 					int flags;
1426 					flags = in6p->in6p_flags;
1427 					if (flags & IN6P_HIGHPORT)
1428 						optval = IPV6_PORTRANGE_HIGH;
1429 					else if (flags & IN6P_LOWPORT)
1430 						optval = IPV6_PORTRANGE_LOW;
1431 					else
1432 						optval = 0;
1433 					break;
1434 				    }
1435 
1436 				case IPV6_PKTINFO:
1437 					optval = OPTBIT(IN6P_PKTINFO);
1438 					break;
1439 
1440 				case IPV6_HOPLIMIT:
1441 					optval = OPTBIT(IN6P_HOPLIMIT);
1442 					break;
1443 
1444 				case IPV6_HOPOPTS:
1445 					optval = OPTBIT(IN6P_HOPOPTS);
1446 					break;
1447 
1448 				case IPV6_DSTOPTS:
1449 					optval = OPTBIT(IN6P_DSTOPTS);
1450 					break;
1451 
1452 				case IPV6_RTHDR:
1453 					optval = OPTBIT(IN6P_RTHDR);
1454 					break;
1455 
1456 				case IPV6_CHECKSUM:
1457 					optval = in6p->in6p_cksum;
1458 					break;
1459 
1460 				case IPV6_FAITH:
1461 					optval = OPTBIT(IN6P_FAITH);
1462 					break;
1463 
1464 #ifndef INET6_BINDV6ONLY
1465 				case IPV6_BINDV6ONLY:
1466 					optval = OPTBIT(IN6P_BINDV6ONLY);
1467 					break;
1468 #endif
1469 				}
1470 				*mtod(m, int *) = optval;
1471 				break;
1472 
1473 			case IPV6_MULTICAST_IF:
1474 			case IPV6_MULTICAST_HOPS:
1475 			case IPV6_MULTICAST_LOOP:
1476 			case IPV6_JOIN_GROUP:
1477 			case IPV6_LEAVE_GROUP:
1478 				error = ip6_getmoptions(optname, in6p->in6p_moptions, mp);
1479 				break;
1480 
1481 #ifdef IPSEC
1482 			case IPV6_IPSEC_POLICY:
1483 			{
1484 				caddr_t req = NULL;
1485 				size_t len = 0;
1486 
1487 				if (m) {
1488 					req = mtod(m, caddr_t);
1489 					len = m->m_len;
1490 				}
1491 				error = ipsec6_get_policy(in6p, req, len, mp);
1492 				break;
1493 			}
1494 #endif /* IPSEC */
1495 
1496 			default:
1497 				error = ENOPROTOOPT;
1498 				break;
1499 			}
1500 			break;
1501 		}
1502 	else {
1503 		error = EINVAL;
1504 		if (op == PRCO_SETOPT && *mp)
1505 			(void)m_free(*mp);
1506 	}
1507 	return(error);
1508 }
1509 
1510 /*
1511  * Set up IP6 options in pcb for insertion in output packets.
1512  * Store in mbuf with pointer in pcbopt, adding pseudo-option
1513  * with destination address if source routed.
1514  */
1515 static int
1516 ip6_pcbopts(pktopt, m, so)
1517 	struct ip6_pktopts **pktopt;
1518 	register struct mbuf *m;
1519 	struct socket *so;
1520 {
1521 	register struct ip6_pktopts *opt = *pktopt;
1522 	int error = 0;
1523 	struct proc *p = curproc;	/* XXX */
1524 	int priv = 0;
1525 
1526 	/* turn off any old options. */
1527 	if (opt) {
1528 		if (opt->ip6po_m)
1529 			(void)m_free(opt->ip6po_m);
1530 	}
1531 	else
1532 		opt = malloc(sizeof(*opt), M_IP6OPT, M_WAITOK);
1533 	*pktopt = 0;
1534 
1535 	if (!m || m->m_len == 0) {
1536 		/*
1537 		 * Only turning off any previous options.
1538 		 */
1539 		if (opt)
1540 			free(opt, M_IP6OPT);
1541 		if (m)
1542 			(void)m_free(m);
1543 		return(0);
1544 	}
1545 
1546 	/*  set options specified by user. */
1547 	if (p && !suser(p->p_ucred, &p->p_acflag))
1548 		priv = 1;
1549 	if ((error = ip6_setpktoptions(m, opt, priv)) != 0) {
1550 		(void)m_free(m);
1551 		return(error);
1552 	}
1553 	*pktopt = opt;
1554 	return(0);
1555 }
1556 
1557 /*
1558  * Set the IP6 multicast options in response to user setsockopt().
1559  */
1560 static int
1561 ip6_setmoptions(optname, im6op, m)
1562 	int optname;
1563 	struct ip6_moptions **im6op;
1564 	struct mbuf *m;
1565 {
1566 	int error = 0;
1567 	u_int loop, ifindex;
1568 	struct ipv6_mreq *mreq;
1569 	struct ifnet *ifp;
1570 	struct ip6_moptions *im6o = *im6op;
1571 	struct route_in6 ro;
1572 	struct sockaddr_in6 *dst;
1573 	struct in6_multi_mship *imm;
1574 	struct proc *p = curproc;	/* XXX */
1575 
1576 	if (im6o == NULL) {
1577 		/*
1578 		 * No multicast option buffer attached to the pcb;
1579 		 * allocate one and initialize to default values.
1580 		 */
1581 		im6o = (struct ip6_moptions *)
1582 			malloc(sizeof(*im6o), M_IPMOPTS, M_WAITOK);
1583 
1584 		if (im6o == NULL)
1585 			return(ENOBUFS);
1586 		*im6op = im6o;
1587 		im6o->im6o_multicast_ifp = NULL;
1588 		im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1589 		im6o->im6o_multicast_loop = IPV6_DEFAULT_MULTICAST_LOOP;
1590 		LIST_INIT(&im6o->im6o_memberships);
1591 	}
1592 
1593 	switch (optname) {
1594 
1595 	case IPV6_MULTICAST_IF:
1596 		/*
1597 		 * Select the interface for outgoing multicast packets.
1598 		 */
1599 		if (m == NULL || m->m_len != sizeof(u_int)) {
1600 			error = EINVAL;
1601 			break;
1602 		}
1603 		ifindex = *(mtod(m, u_int *));
1604 		if (ifindex < 0 || if_index < ifindex) {
1605 			error = ENXIO;	/* XXX EINVAL? */
1606 			break;
1607 		}
1608 		ifp = ifindex2ifnet[ifindex];
1609 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1610 			error = EADDRNOTAVAIL;
1611 			break;
1612 		}
1613 		im6o->im6o_multicast_ifp = ifp;
1614 		break;
1615 
1616 	case IPV6_MULTICAST_HOPS:
1617 	    {
1618 		/*
1619 		 * Set the IP6 hoplimit for outgoing multicast packets.
1620 		 */
1621 		int optval;
1622 		if (m == NULL || m->m_len != sizeof(int)) {
1623 			error = EINVAL;
1624 			break;
1625 		}
1626 		optval = *(mtod(m, u_int *));
1627 		if (optval < -1 || optval >= 256)
1628 			error = EINVAL;
1629 		else if (optval == -1)
1630 			im6o->im6o_multicast_hlim = ip6_defmcasthlim;
1631 		else
1632 			im6o->im6o_multicast_hlim = optval;
1633 		break;
1634 	    }
1635 
1636 	case IPV6_MULTICAST_LOOP:
1637 		/*
1638 		 * Set the loopback flag for outgoing multicast packets.
1639 		 * Must be zero or one.
1640 		 */
1641 		if (m == NULL || m->m_len != sizeof(u_int) ||
1642 		   (loop = *(mtod(m, u_int *))) > 1) {
1643 			error = EINVAL;
1644 			break;
1645 		}
1646 		im6o->im6o_multicast_loop = loop;
1647 		break;
1648 
1649 	case IPV6_JOIN_GROUP:
1650 		/*
1651 		 * Add a multicast group membership.
1652 		 * Group must be a valid IP6 multicast address.
1653 		 */
1654 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1655 			error = EINVAL;
1656 			break;
1657 		}
1658 		mreq = mtod(m, struct ipv6_mreq *);
1659 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1660 			/*
1661 			 * We use the unspecified address to specify to accept
1662 			 * all multicast addresses. Only super user is allowed
1663 			 * to do this.
1664 			 */
1665 			if (suser(p->p_ucred, &p->p_acflag)) {
1666 				error = EACCES;
1667 				break;
1668 			}
1669 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1670 			error = EINVAL;
1671 			break;
1672 		}
1673 
1674 		/*
1675 		 * If the interface is specified, validate it.
1676 		 */
1677 		if (mreq->ipv6mr_interface < 0
1678 		 || if_index < mreq->ipv6mr_interface) {
1679 			error = ENXIO;	/* XXX EINVAL? */
1680 			break;
1681 		}
1682 		/*
1683 		 * If no interface was explicitly specified, choose an
1684 		 * appropriate one according to the given multicast address.
1685 		 */
1686 		if (mreq->ipv6mr_interface == 0) {
1687 			/*
1688 			 * If the multicast address is in node-local scope,
1689 			 * the interface should be a loopback interface.
1690 			 * Otherwise, look up the routing table for the
1691 			 * address, and choose the outgoing interface.
1692 			 *   XXX: is it a good approach?
1693 			 */
1694 			if (IN6_IS_ADDR_MC_NODELOCAL(&mreq->ipv6mr_multiaddr)) {
1695 				ifp = &loif[0];
1696 			}
1697 			else {
1698 				ro.ro_rt = NULL;
1699 				dst = (struct sockaddr_in6 *)&ro.ro_dst;
1700 				bzero(dst, sizeof(*dst));
1701 				dst->sin6_len = sizeof(struct sockaddr_in6);
1702 				dst->sin6_family = AF_INET6;
1703 				dst->sin6_addr = mreq->ipv6mr_multiaddr;
1704 				rtalloc((struct route *)&ro);
1705 				if (ro.ro_rt == NULL) {
1706 					error = EADDRNOTAVAIL;
1707 					break;
1708 				}
1709 				ifp = ro.ro_rt->rt_ifp;
1710 				rtfree(ro.ro_rt);
1711 			}
1712 		} else
1713 			ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1714 
1715 		/*
1716 		 * See if we found an interface, and confirm that it
1717 		 * supports multicast
1718 		 */
1719 		if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0) {
1720 			error = EADDRNOTAVAIL;
1721 			break;
1722 		}
1723 		/*
1724 		 * Put interface index into the multicast address,
1725 		 * if the address has link-local scope.
1726 		 */
1727 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1728 			mreq->ipv6mr_multiaddr.s6_addr16[1]
1729 				= htons(mreq->ipv6mr_interface);
1730 		}
1731 		/*
1732 		 * See if the membership already exists.
1733 		 */
1734 		for (imm = im6o->im6o_memberships.lh_first;
1735 		     imm != NULL; imm = imm->i6mm_chain.le_next)
1736 			if (imm->i6mm_maddr->in6m_ifp == ifp &&
1737 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1738 					       &mreq->ipv6mr_multiaddr))
1739 				break;
1740 		if (imm != NULL) {
1741 			error = EADDRINUSE;
1742 			break;
1743 		}
1744 		/*
1745 		 * Everything looks good; add a new record to the multicast
1746 		 * address list for the given interface.
1747 		 */
1748 		imm = malloc(sizeof(*imm), M_IPMADDR, M_WAITOK);
1749 		if (imm == NULL) {
1750 			error = ENOBUFS;
1751 			break;
1752 		}
1753 		if ((imm->i6mm_maddr =
1754 		     in6_addmulti(&mreq->ipv6mr_multiaddr, ifp, &error)) == NULL) {
1755 			free(imm, M_IPMADDR);
1756 			break;
1757 		}
1758 		LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1759 		break;
1760 
1761 	case IPV6_LEAVE_GROUP:
1762 		/*
1763 		 * Drop a multicast group membership.
1764 		 * Group must be a valid IP6 multicast address.
1765 		 */
1766 		if (m == NULL || m->m_len != sizeof(struct ipv6_mreq)) {
1767 			error = EINVAL;
1768 			break;
1769 		}
1770 		mreq = mtod(m, struct ipv6_mreq *);
1771 		if (IN6_IS_ADDR_UNSPECIFIED(&mreq->ipv6mr_multiaddr)) {
1772 			if (suser(p->p_ucred, &p->p_acflag)) {
1773 				error = EACCES;
1774 				break;
1775 			}
1776 		} else if (!IN6_IS_ADDR_MULTICAST(&mreq->ipv6mr_multiaddr)) {
1777 			error = EINVAL;
1778 			break;
1779 		}
1780 		/*
1781 		 * If an interface address was specified, get a pointer
1782 		 * to its ifnet structure.
1783 		 */
1784 		if (mreq->ipv6mr_interface < 0
1785 		 || if_index < mreq->ipv6mr_interface) {
1786 			error = ENXIO;	/* XXX EINVAL? */
1787 			break;
1788 		}
1789 		ifp = ifindex2ifnet[mreq->ipv6mr_interface];
1790 		/*
1791 		 * Put interface index into the multicast address,
1792 		 * if the address has link-local scope.
1793 		 */
1794 		if (IN6_IS_ADDR_MC_LINKLOCAL(&mreq->ipv6mr_multiaddr)) {
1795 			mreq->ipv6mr_multiaddr.s6_addr16[1]
1796 				= htons(mreq->ipv6mr_interface);
1797 		}
1798 		/*
1799 		 * Find the membership in the membership list.
1800 		 */
1801 		for (imm = im6o->im6o_memberships.lh_first;
1802 		     imm != NULL; imm = imm->i6mm_chain.le_next) {
1803 			if ((ifp == NULL ||
1804 			     imm->i6mm_maddr->in6m_ifp == ifp) &&
1805 			    IN6_ARE_ADDR_EQUAL(&imm->i6mm_maddr->in6m_addr,
1806 					       &mreq->ipv6mr_multiaddr))
1807 				break;
1808 		}
1809 		if (imm == NULL) {
1810 			/* Unable to resolve interface */
1811 			error = EADDRNOTAVAIL;
1812 			break;
1813 		}
1814 		/*
1815 		 * Give up the multicast address record to which the
1816 		 * membership points.
1817 		 */
1818 		LIST_REMOVE(imm, i6mm_chain);
1819 		in6_delmulti(imm->i6mm_maddr);
1820 		free(imm, M_IPMADDR);
1821 		break;
1822 
1823 	default:
1824 		error = EOPNOTSUPP;
1825 		break;
1826 	}
1827 
1828 	/*
1829 	 * If all options have default values, no need to keep the mbuf.
1830 	 */
1831 	if (im6o->im6o_multicast_ifp == NULL &&
1832 	    im6o->im6o_multicast_hlim == ip6_defmcasthlim &&
1833 	    im6o->im6o_multicast_loop == IPV6_DEFAULT_MULTICAST_LOOP &&
1834 	    im6o->im6o_memberships.lh_first == NULL) {
1835 		free(*im6op, M_IPMOPTS);
1836 		*im6op = NULL;
1837 	}
1838 
1839 	return(error);
1840 }
1841 
1842 /*
1843  * Return the IP6 multicast options in response to user getsockopt().
1844  */
1845 static int
1846 ip6_getmoptions(optname, im6o, mp)
1847 	int optname;
1848 	register struct ip6_moptions *im6o;
1849 	register struct mbuf **mp;
1850 {
1851 	u_int *hlim, *loop, *ifindex;
1852 
1853 	*mp = m_get(M_WAIT, MT_SOOPTS);
1854 
1855 	switch (optname) {
1856 
1857 	case IPV6_MULTICAST_IF:
1858 		ifindex = mtod(*mp, u_int *);
1859 		(*mp)->m_len = sizeof(u_int);
1860 		if (im6o == NULL || im6o->im6o_multicast_ifp == NULL)
1861 			*ifindex = 0;
1862 		else
1863 			*ifindex = im6o->im6o_multicast_ifp->if_index;
1864 		return(0);
1865 
1866 	case IPV6_MULTICAST_HOPS:
1867 		hlim = mtod(*mp, u_int *);
1868 		(*mp)->m_len = sizeof(u_int);
1869 		if (im6o == NULL)
1870 			*hlim = ip6_defmcasthlim;
1871 		else
1872 			*hlim = im6o->im6o_multicast_hlim;
1873 		return(0);
1874 
1875 	case IPV6_MULTICAST_LOOP:
1876 		loop = mtod(*mp, u_int *);
1877 		(*mp)->m_len = sizeof(u_int);
1878 		if (im6o == NULL)
1879 			*loop = ip6_defmcasthlim;
1880 		else
1881 			*loop = im6o->im6o_multicast_loop;
1882 		return(0);
1883 
1884 	default:
1885 		return(EOPNOTSUPP);
1886 	}
1887 }
1888 
1889 /*
1890  * Discard the IP6 multicast options.
1891  */
1892 void
1893 ip6_freemoptions(im6o)
1894 	register struct ip6_moptions *im6o;
1895 {
1896 	struct in6_multi_mship *imm;
1897 
1898 	if (im6o == NULL)
1899 		return;
1900 
1901 	while ((imm = im6o->im6o_memberships.lh_first) != NULL) {
1902 		LIST_REMOVE(imm, i6mm_chain);
1903 		if (imm->i6mm_maddr)
1904 			in6_delmulti(imm->i6mm_maddr);
1905 		free(imm, M_IPMADDR);
1906 	}
1907 	free(im6o, M_IPMOPTS);
1908 }
1909 
1910 /*
1911  * Set IPv6 outgoing packet options based on advanced API.
1912  */
1913 int
1914 ip6_setpktoptions(control, opt, priv)
1915 	struct mbuf *control;
1916 	struct ip6_pktopts *opt;
1917 	int priv;
1918 {
1919 	register struct cmsghdr *cm = 0;
1920 
1921 	if (control == 0 || opt == 0)
1922 		return(EINVAL);
1923 
1924 	bzero(opt, sizeof(*opt));
1925 	opt->ip6po_hlim = -1; /* -1 means to use default hop limit */
1926 
1927 	/*
1928 	 * XXX: Currently, we assume all the optional information is stored
1929 	 * in a single mbuf.
1930 	 */
1931 	if (control->m_next)
1932 		return(EINVAL);
1933 
1934 	opt->ip6po_m = control;
1935 
1936 	for (; control->m_len; control->m_data += CMSG_ALIGN(cm->cmsg_len),
1937 		     control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
1938 		cm = mtod(control, struct cmsghdr *);
1939 		if (cm->cmsg_len == 0 || cm->cmsg_len > control->m_len)
1940 			return(EINVAL);
1941 		if (cm->cmsg_level != IPPROTO_IPV6)
1942 			continue;
1943 
1944 		switch(cm->cmsg_type) {
1945 		case IPV6_PKTINFO:
1946 			if (cm->cmsg_len != CMSG_LEN(sizeof(struct in6_pktinfo)))
1947 				return(EINVAL);
1948 			opt->ip6po_pktinfo = (struct in6_pktinfo *)CMSG_DATA(cm);
1949 			if (opt->ip6po_pktinfo->ipi6_ifindex &&
1950 			    IN6_IS_ADDR_LINKLOCAL(&opt->ip6po_pktinfo->ipi6_addr))
1951 				opt->ip6po_pktinfo->ipi6_addr.s6_addr16[1] =
1952 					htons(opt->ip6po_pktinfo->ipi6_ifindex);
1953 
1954 			if (opt->ip6po_pktinfo->ipi6_ifindex > if_index
1955 			 || opt->ip6po_pktinfo->ipi6_ifindex < 0) {
1956 				return(ENXIO);
1957 			}
1958 
1959 			if (!IN6_IS_ADDR_UNSPECIFIED(&opt->ip6po_pktinfo->ipi6_addr)) {
1960 				struct ifaddr *ia;
1961 				struct sockaddr_in6 sin6;
1962 
1963 				bzero(&sin6, sizeof(sin6));
1964 				sin6.sin6_len = sizeof(sin6);
1965 				sin6.sin6_family = AF_INET6;
1966 				sin6.sin6_addr =
1967 					opt->ip6po_pktinfo->ipi6_addr;
1968 				ia = ifa_ifwithaddr(sin6tosa(&sin6));
1969 				if (ia == NULL ||
1970 				    (opt->ip6po_pktinfo->ipi6_ifindex &&
1971 				     (ia->ifa_ifp->if_index !=
1972 				      opt->ip6po_pktinfo->ipi6_ifindex))) {
1973 					return(EADDRNOTAVAIL);
1974 				}
1975 				/*
1976 				 * Check if the requested source address is
1977 				 * indeed a unicast address assigned to the
1978 				 * node.
1979 				 */
1980 				if (IN6_IS_ADDR_MULTICAST(&opt->ip6po_pktinfo->ipi6_addr))
1981 					return(EADDRNOTAVAIL);
1982 			}
1983 			break;
1984 
1985 		case IPV6_HOPLIMIT:
1986 			if (cm->cmsg_len != CMSG_LEN(sizeof(int)))
1987 				return(EINVAL);
1988 
1989 			opt->ip6po_hlim = *(int *)CMSG_DATA(cm);
1990 			if (opt->ip6po_hlim < -1 || opt->ip6po_hlim > 255)
1991 				return(EINVAL);
1992 			break;
1993 
1994 		case IPV6_NEXTHOP:
1995 			if (!priv)
1996 				return(EPERM);
1997 			if (cm->cmsg_len < sizeof(u_char) ||
1998 			    cm->cmsg_len < CMSG_LEN(*CMSG_DATA(cm)))
1999 				return(EINVAL);
2000 
2001 			opt->ip6po_nexthop = (struct sockaddr *)CMSG_DATA(cm);
2002 
2003 			break;
2004 
2005 		case IPV6_HOPOPTS:
2006 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_hbh)))
2007 				return(EINVAL);
2008 			opt->ip6po_hbh = (struct ip6_hbh *)CMSG_DATA(cm);
2009 			if (cm->cmsg_len !=
2010 			    CMSG_LEN((opt->ip6po_hbh->ip6h_len + 1) << 3))
2011 				return(EINVAL);
2012 			break;
2013 
2014 		case IPV6_DSTOPTS:
2015 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_dest)))
2016 				return(EINVAL);
2017 
2018 			/*
2019 			 * If there is no routing header yet, the destination
2020 			 * options header should be put on the 1st part.
2021 			 * Otherwise, the header should be on the 2nd part.
2022 			 * (See RFC 2460, section 4.1)
2023 			 */
2024 			if (opt->ip6po_rthdr == NULL) {
2025 				opt->ip6po_dest1 =
2026 					(struct ip6_dest *)CMSG_DATA(cm);
2027 				if (cm->cmsg_len !=
2028 				    CMSG_LEN((opt->ip6po_dest1->ip6d_len + 1)
2029 					     << 3))
2030 					return(EINVAL);
2031 			}
2032 			else {
2033 				opt->ip6po_dest2 =
2034 					(struct ip6_dest *)CMSG_DATA(cm);
2035 				if (cm->cmsg_len !=
2036 				    CMSG_LEN((opt->ip6po_dest2->ip6d_len + 1)
2037 					     << 3))
2038 					return(EINVAL);
2039 			}
2040 			break;
2041 
2042 		case IPV6_RTHDR:
2043 			if (cm->cmsg_len < CMSG_LEN(sizeof(struct ip6_rthdr)))
2044 				return(EINVAL);
2045 			opt->ip6po_rthdr = (struct ip6_rthdr *)CMSG_DATA(cm);
2046 			if (cm->cmsg_len !=
2047 			    CMSG_LEN((opt->ip6po_rthdr->ip6r_len + 1) << 3))
2048 				return(EINVAL);
2049 			switch(opt->ip6po_rthdr->ip6r_type) {
2050 			case IPV6_RTHDR_TYPE_0:
2051 				if (opt->ip6po_rthdr->ip6r_segleft == 0)
2052 					return(EINVAL);
2053 				break;
2054 			default:
2055 				return(EINVAL);
2056 			}
2057 			break;
2058 
2059 		default:
2060 			return(ENOPROTOOPT);
2061 		}
2062 	}
2063 
2064 	return(0);
2065 }
2066 
2067 /*
2068  * Routine called from ip6_output() to loop back a copy of an IP6 multicast
2069  * packet to the input queue of a specified interface.  Note that this
2070  * calls the output routine of the loopback "driver", but with an interface
2071  * pointer that might NOT be &loif -- easier than replicating that code here.
2072  */
2073 void
2074 ip6_mloopback(ifp, m, dst)
2075 	struct ifnet *ifp;
2076 	register struct mbuf *m;
2077 	register struct sockaddr_in6 *dst;
2078 {
2079 	struct	mbuf *copym;
2080 
2081 	copym = m_copy(m, 0, M_COPYALL);
2082 	if (copym != NULL)
2083 		(void)looutput(ifp, copym, (struct sockaddr *)dst, NULL);
2084 }
2085 
2086 /*
2087  * Chop IPv6 header off from the payload.
2088  */
2089 static int
2090 ip6_splithdr(m, exthdrs)
2091 	struct mbuf *m;
2092 	struct ip6_exthdrs *exthdrs;
2093 {
2094 	struct mbuf *mh;
2095 	struct ip6_hdr *ip6;
2096 
2097 	ip6 = mtod(m, struct ip6_hdr *);
2098 	if (m->m_len > sizeof(*ip6)) {
2099 		MGETHDR(mh, M_DONTWAIT, MT_HEADER);
2100 		if (mh == 0) {
2101 			m_freem(m);
2102 			return ENOBUFS;
2103 		}
2104 		M_COPY_PKTHDR(mh, m);
2105 		MH_ALIGN(mh, sizeof(*ip6));
2106 		m->m_flags &= ~M_PKTHDR;
2107 		m->m_len -= sizeof(*ip6);
2108 		m->m_data += sizeof(*ip6);
2109 		mh->m_next = m;
2110 		m = mh;
2111 		m->m_len = sizeof(*ip6);
2112 		bcopy((caddr_t)ip6, mtod(m, caddr_t), sizeof(*ip6));
2113 	}
2114 	exthdrs->ip6e_ip6 = m;
2115 	return 0;
2116 }
2117 
2118 /*
2119  * Compute IPv6 extension header length.
2120  */
2121 int
2122 ip6_optlen(in6p)
2123 	struct in6pcb *in6p;
2124 {
2125 	int len;
2126 
2127 	if (!in6p->in6p_outputopts)
2128 		return 0;
2129 
2130 	len = 0;
2131 #define elen(x) \
2132     (((struct ip6_ext *)(x)) ? (((struct ip6_ext *)(x))->ip6e_len + 1) << 3 : 0)
2133 
2134 	len += elen(in6p->in6p_outputopts->ip6po_hbh);
2135 	len += elen(in6p->in6p_outputopts->ip6po_dest1);
2136 	len += elen(in6p->in6p_outputopts->ip6po_rthdr);
2137 	len += elen(in6p->in6p_outputopts->ip6po_dest2);
2138 	return len;
2139 #undef elen
2140 }
2141